Description
People rate how convincing it is that a property of some categories also holds for another, revealing how similarity, coverage and causal knowledge drive generalisation.
Participants rate arguments that project a novel property from premise categories to a conclusion category, such as 'cows have sesamoid bones, so dogs do'. The similarity-coverage model predicts strength from the similarity of premises to conclusion plus coverage: how well the premises span the smallest superordinate that contains them all. Diverse premises give more coverage, atypical premises little. The type of property changes which similarity counts (anatomical versus behavioural), causal links in the premises raise strength, and adding a distant premise can lower it. Experts often reason from causal knowledge and ignore diversity. The method is rigorous because ratings are compared against quantitative model predictions under systematic manipulations of similarity, typicality, diversity and causal links.
Attributes
- Discipline
- Behavioral Science & Economics
- Loss type
- Unmigrated
- Data
- Quantitative
- Effort
- Days
Steps
- Choose premise and conclusion categories with known similarity and typicality (from prior ratings).
- Attach a blank property (unfamiliar to participants) or a domain-specific one, depending on the question.
- Build argument sets that vary premise diversity, typicality and causal relations while holding the conclusion constant.
- Present arguments in random order and collect strength or likelihood ratings on a fixed scale.
- Compare ratings with similarity and coverage predictions, and test diversity, typicality and nonmonotonicity effects.
- Repeat with expert and novice groups to see whether causal knowledge overrides similarity.
Sources
- Handbook of Perception and Cognition (Bly & Rumelhart, 1999)
- Osherson, D. N., Smith, E. E., Wilkie, O., Lopez, A., & Shafir, E. (1990). Category-based induction. Psychological Review, 97
Also known as
- Category-based induction argument-strength rating (similarity-coverage model)
- Similarity-coverage model
- Property projection task
Categories
- Analytics & Behaviour
- Behavioral Science & Economics